In forest ecosystems litter is usually assessed in terms of the average amount produced by the canopy. In scattered tree ecosystems this approach is problematic because the canopy is discontinuous and the spatial arrangement of litter highly variable. We addressed this problem by quantifying the spatial variation in litter load and litter composition associated with individual trees in a Eucalyptus melliodora-Eucalyptus blakelyi woodland. Litter was sampled under crowns and in grassland adjacent to 10 E. blakelyi and 10 E. melliodora trees ranging in diameter at breast height (dbh) from 14 to 129 cm. A total of 302 L samples were collected from these trees, at distances ranging from 0 to 42 m from main stem. The sampled litter loads ranged from 0.02 to 109.3 t ha -1 and were significantly higher under tree crowns than in grassland for litter and each component of litter (leaves, bark, fine twigs, coarse twigs). In particular, the mean litter load under tree crowns (12.5 t ha -1) was an order of magnitude higher than the mean litter load in grassland (1.27 t ha-1). There was a significant (P = 0.0103) positive relationship between mean litter load under the tree crown and dbh, indicating larger trees produced more litter per unit area of ground than smaller trees. Generalized Linear Modelling produced highly significant (P < 0.0001) models predicting the spatial variation in litter load and litter composition in terms of distance from main stem and dbh. Our models demonstrate gradients in litter load and composition under tree crowns. These gradients were most pronounced for the large trees in our study. The disproportionate input of litter and variety of litter components associated with large trees in our study supports their keystone role in scattered tree ecosystems and highlights the need to maintain these structures in agricultural landscapes.